Human oversight policy

Purpose

This policy describes the mandatory human oversight step in Rolerise resume matching. The matching engine scores, ranks, and explains candidate fit. It does not reject candidates, close applications, or send adverse-action notices.

What the AI may do

The system may parse job requirements, validate readable CVs, detect likely duplicates, score valid candidates against the role, and produce a ranked shortlist with evidence quotes from the source resume.

Scores and explanations are decision-support for a recruiter. They are not hiring decisions.

What the AI must not do

The AI must not auto-reject, auto-decline, or notify a candidate that they are unsuccessful. No candidate is removed from process solely because of a model score.

Recruiters remain responsible for shortlist review, interview decisions, and all communications with candidates.

Required recruiter step

Before any rejection, hold, or offer, a human reviewer must open the shortlist (or the relevant bucket), read the score breakdown and evidence, and record the action. The product is designed so this review cannot be skipped for production tenants.